A Technical Review on Unsupervised Learning of Graph and Hypergraph Pattern Analysis

Graph and hypergraph matching are fundamental problems in pattern analysis problems. They are applied to various tasks requiring 2D and 3D feature matching, such as image alignment, 3D reconstruction, and object or action recognition. Graph pattern analysis considers pairwise constraints that usually encode geometric and appearance associations between local features. On the other hand, hypergraph matching incorporates higher-order relations computed over sets of features, which could capture both geometric and appearance information. Therefore, using higher-order constraints enables matching that is more robust (or even invariant) to changes in scale, non-rigid deformations, and outliers. Many objects or other entities such as gesture recognition and human activities in the spatiotemporal domain can be signified by graphs with local information on nodes and more global information on edges or hyperedges. In this research, and essential review have been done on the unsupervised methods to explore and communicate meta-analytic data and results with a large number of novel graphs proposed quite recently.